The canonical genetic code has been reported both to be error minimizing and to show stereochemical associations between coding triplets and binding sites. In order to test whether these two properties are unexpectedly overlapping, we generated 200,000 randomized genetic codes using each of five randomization schemes, with and without randomization of stop codons. Comparison of the code error (difference in polar requirement for single-nucleotide codon interchanges) with the coding triplet concentrations in RNA binding sites for eight amino acids shows that these properties are independent and uncorrelated. Thus, one is not the result of the other, and error minimization and triplet associations probably arose independently during the history of the genetic code. We explicitly show that prior fixation of a stereochemical core is consistent with an effective later minimization of error.